10 Chemometric Analysis of Raman and IR Spectra of Natural Dyes
289
Apart from single in situ measurements the mapping of biological tissues can
be performed followed by the chemometric analysis [60]. First, the spectra are collected from the surface of sample and then data are processed to get the chemical
image of the sample. the most popular method is the integration of the spectral intensity signal around the spectral position of characteristic vibrational bands of the
pigments but also hierarchical cluster analysis (hCA) is often used. Finally the 2d
or 3d chemical image of the distribution and relative concentration of the analyte is
obtained. It should be remembered that this technique often requires to define the so
called marker band or characteristic spectral range of substances under study, what
is very often problematic. the Raman (or IR) bands may be useful as markers for
Raman (or IR) sample mapping, as long as one can avoid band overlap hindering a
clear separation of other biological components e.g. lignin, cellulose, chlorophyll
[9, 27, 28, 61] or proteins [9].
In case of the problem with defining of marker bands, also the chemometry techniques could be used. only the few examples are briefly described below.
Principal component analysis (PCA) is often used for spectroscopic analysis.
the main idea of that method is an extraction of data and in consequence, reduction of the number of variables that contain maximal variance. In some cases, the
number of different components could be determined with this method [62]. the
similar application has linear discriminant analysis (LdA) which is also applied to
data classification.
Cluster analysis (CA) groups data according to their similarities. the most popular types of CA are hierarchical CA (hCA) and K-means CA. Finally, one gets a
dendrogram, where the dependence between data is represented by a branches of
tree or, in case of imaging techniques, certain number of areas which shows a distribution of various clusters in the surface [62, 63].
the interesting method of computational analysis of spectra is 2d-correlation
spectroscopy, which originates from 2d NmR spectroscopy and was developed by
Isao Noda in 80’s. the idea of the 2d correlation analysis is to subject a sample to an
external perturbation while all other parameters of the system are kept at the same
value. Spectra of the initial sample and subjected to the perturbation are correlated
one to another. the nature of perturbation can be different e.g. temperature, pressure, ph or changes of the chemical composition of the system. that computational
method could be a convenient tool for detection and analysis even minor changes
of the spectra. more information about this method one can find elsewhere [64].
the background for quantitative analysis is the fact that Raman scattering is
directly proportional to the concentration of the vibrating species [51]. In case of
IR techniques, the absorption is proportional to the concentration of the analyte according to the Lambert-Beer law. Both methods can be successfully supported in
quantification analysis with the use of chemometric methodes.
In case of IR spectroscopy chemometric support has been widely used mainly
in near infrared range, especially during in situ measurements. this is because NIR
measurements might be influenced by a high water content of a sample, which
strongly disturbs the analysis, especially from the water bands which are additionally enhanced by their multiplication (overtones) [10].
289
Apart from single in situ measurements the mapping of biological tissues can
be performed followed by the chemometric analysis [60]. First, the spectra are collected from the surface of sample and then data are processed to get the chemical
image of the sample. the most popular method is the integration of the spectral intensity signal around the spectral position of characteristic vibrational bands of the
pigments but also hierarchical cluster analysis (hCA) is often used. Finally the 2d
or 3d chemical image of the distribution and relative concentration of the analyte is
obtained. It should be remembered that this technique often requires to define the so
called marker band or characteristic spectral range of substances under study, what
is very often problematic. the Raman (or IR) bands may be useful as markers for
Raman (or IR) sample mapping, as long as one can avoid band overlap hindering a
clear separation of other biological components e.g. lignin, cellulose, chlorophyll
[9, 27, 28, 61] or proteins [9].
In case of the problem with defining of marker bands, also the chemometry techniques could be used. only the few examples are briefly described below.
Principal component analysis (PCA) is often used for spectroscopic analysis.
the main idea of that method is an extraction of data and in consequence, reduction of the number of variables that contain maximal variance. In some cases, the
number of different components could be determined with this method [62]. the
similar application has linear discriminant analysis (LdA) which is also applied to
data classification.
Cluster analysis (CA) groups data according to their similarities. the most popular types of CA are hierarchical CA (hCA) and K-means CA. Finally, one gets a
dendrogram, where the dependence between data is represented by a branches of
tree or, in case of imaging techniques, certain number of areas which shows a distribution of various clusters in the surface [62, 63].
the interesting method of computational analysis of spectra is 2d-correlation
spectroscopy, which originates from 2d NmR spectroscopy and was developed by
Isao Noda in 80’s. the idea of the 2d correlation analysis is to subject a sample to an
external perturbation while all other parameters of the system are kept at the same
value. Spectra of the initial sample and subjected to the perturbation are correlated
one to another. the nature of perturbation can be different e.g. temperature, pressure, ph or changes of the chemical composition of the system. that computational
method could be a convenient tool for detection and analysis even minor changes
of the spectra. more information about this method one can find elsewhere [64].
the background for quantitative analysis is the fact that Raman scattering is
directly proportional to the concentration of the vibrating species [51]. In case of
IR techniques, the absorption is proportional to the concentration of the analyte according to the Lambert-Beer law. Both methods can be successfully supported in
quantification analysis with the use of chemometric methodes.
In case of IR spectroscopy chemometric support has been widely used mainly
in near infrared range, especially during in situ measurements. this is because NIR
measurements might be influenced by a high water content of a sample, which
strongly disturbs the analysis, especially from the water bands which are additionally enhanced by their multiplication (overtones) [10].